S2E1: Top Talent Invests In Transferrable Skills | Yash Tekriwal

Welcome back everyone to the GTM Engineer Podcast, Season 2, from hype to operating system. So we know all that GTM engineering is here to stay, so we are here to make some sense out of it. And I think there's no better person to talk about the foundation of this discipline than today's guest. If you've learned Clay in the past couple of years, there's a good chance that Yash Tikriwal taught you. He's the head of education at Clay, but the title pretty much undersells it. So before running education, Yash was actually the very first Clay's founding GTM engineer. He built out their core sales processes, the demo tables, and created the custom POC model that helped Clay hit back-to-back years of 6 to 10x revenue growth. And, uh, here's what makes Yash's perspectives unique. He's a former high school teacher and a 2x founder and a digital nomad that turned, uh, GTM architect. So he brings all of this, uh, incredible mix, uh, of, uh, experiences and skills into how he thinks about, uh, enabling the next generation of GTM teams, uh, at scale. So, um, yeah, what's his take on what we're all doing? People call, call it automation, but, uh, what you're really doing is, uh, teaching people how to program their business without Python. And I think that reframe is, uh, super powerful. It's, uh, again, we're shifting from, uh, just clicking buttons on a screen to actual computational thinking, right? And that's why Clay University and his cohort-based programs are producing some of the sharpest, uh, GTM engineers in the game. So today we're gonna cover a lot of ground. We'll map out the key milestone and stages of learning for GTM engineers from beginner to advanced. We'll dig into the biggest obstacles people hit when learning this discipline, and, uh, we'll also chat, uh, some of the latest and greatest, uh, Play's features. So Yash, I'm so pumped to have you on the show.

Yeah, I'm stoked to be here. That is the most flattering intro I've ever gotten from an individual, so thank you for that. I don't know if I'm actually that cool, but happy to hear it.

I think you are. I mean, like, I probably am not the only person who would say that, but all right. Yash, like, GTM engineering is not new anymore, you know, like, uh, since you guys coined the term, uh, back in 2023, how have you seen it, uh, changing, especially in the last year?

Yeah, it's a, it's a meaty question, right? There's lots of ways in which GTM engineering is continuing to change. There's also still the debates about whether or not it's just a fancy new title or a fad. I actually think the thing that I read most recently that I most deeply agree with is actually a little blurb from Andreessen on the evolution of all these engineering job titles, which we've been talking a lot about. I think actually what's interesting to look at is people thought prompt engineer was gonna be a job title for a while. I don't think you see anyone hiring prompt engineers, right? Prompt engineering is actually just an element of GTM engineering and content engineering and marketing engineering, because building prompts to build systems, to build frameworks that automate what you were previously unable to automate without code is the big unlock. Now, where that's going to lead the industry, only time will tell. I think the best parallel is to look at the evolution of software engineering actually over the years. If you look at before we called them software engineers, We actually used to call all the people writing basic IT and admin support, right? And then that got changed to a sexier, like, software engineering job title because you were actually creating, you weren't just maintaining. And then even software engineer got sort of bifurcated into front-end, web dev, back-end, because the discipline grew and therefore the specialties needed to grow as well. And then even today you have the evolution that everyone loves of the forward deployed engineer that Palantir really coined. Right? And now we're taking these engineering skills and putting them in front of customers. And I would be willing to bet you a lot of money that that evolution will only continue to develop as more and more people get skilled at being able to do things that previously only a small group of people were able to do. The possibilities exponentially multiply, cuz now you have a bunch of people who are both able to be customer-facing and have dev experience, whereas 15 years ago you didn't have that. It was one or the other.

Very telling. And let's say that GTM engineer disappeared tomorrow. What would you miss the most about it? You know, like what would it be, you know, like most desperate about without all this new series of opportunities and capabilities that we all, that are now so democratized?

Yeah, it's a strange question, right? Because like, you know, I am just, how would GTM engineering disappear, right? Like you could take the title away, but people are still gonna be doing the work with AI. So like if we take, all the capabilities that make GTM engineering possible away, then that just means we've rewinded the clock to 2019. And then the thing that I would miss the most is the ability to do the work that I actually enjoy. The thing that I would hate the most is the reintroduction of me pouring through multi-page documents and Slacks to try and understand what the point someone was trying to get across is. I think now I love being able to hit AI to quickly summarize quick points. I love being able to go through my Slack. Without worrying about having to get every single detail. I love being able to aggregate all the information that I need. I don't miss manually Google searching people's names 17 different ways on LinkedIn to figure out what the most personal thing I could find about them is, and then sitting down to wonder what 6 words I'm gonna pull out of that research to include in an outbound email to them. It was fun the first couple of times. It was rewarding the first few times someone responded, and then after, you know, Email 10, you're like, I get it. I can do this. It's fun. Thank you.

Move on. No more grunt work, please.

Yeah.

Yeah. No more grunt work. Is that really the reality? And, and where does the, where is the hype overselling this whole space? Are there still steps that you see people doing manually, uh, that you think are gonna go away in the future?

Yeah, it's a very good question. I would say there's tons of hype. I also sort of think about this as like, you know, the way elections happen, right? Like no one wins the presidency on, you know, make America 20% better. Uh, but in practice, that's really how things actually evolve, right? A lot of people are like, AI will automate all the grunt work. In reality, what I think AI is doing, which is still a step change forward, is it is moving the grunt work up one layer of abstraction. So it's slightly less gruntish, and I can give an example in a second, and it's also further automating the amount of time spent on that section of work. So as an example, we just implemented a custom agent that takes daily updates that my entire team puts into Slack and then automatically aggregates not only those daily updates, but all of the other Slack, Notion updates they've done in that week and puts them into a weekly update format that previously everyone had to fill out manually. This isn't even GTM engineering. This is just business team ops engineering. Right before people had to manually input that work in there, it was like another 15 to 20 minutes of their day. It's the most soul-sucking work of all time because you've already put in your daily updates for the last 5 days. Now you have to go figure out what the consolidation is, make it really clean and concise, and put it out there because the rest of the team needs to benefit from it. Now, does that make all the grunt work go away? No, we've removed the grunt work of I have to go write. All those things. What we have added now is because everyone is reliably, consistently, and at a high quality getting those updates in there, I'm asking everyone to still read everyone else's updates. 'Cause I know that no one was doing it before and I still know that, you know, people are not going to benefit from it as much because it's being done haphazardly. But me reading through everyone else's updates is still a little bit of that maybe grunt work. It's not the same level of grunt as me having to go regurgitate the same information again. And it's also an added layer of fun and creative. Like what we have noticed in the last few team meetings is that we're actually having more interconnected team follow-ups because we're actually talking about the things that we're working on as opposed to trying to catch up on what was done over the week.

So that's a really interesting example. I'm curious, is there a process by which y'all prioritize what you're going to, uh, tune, uh, these capabilities on?

Yeah. It's a good question. So I think if I'm understanding the question correctly, you're asking, is there a way in which we prioritize what things we decide to automate, what things we spend time on and what things—

Yeah. Why did you decide to prioritize summarizing updates versus something else? We, you know, there's so many things now we can sort of hook up, automate, add AI, scrape, write. It can, in a sense, be overwhelming. Is there like a framework?

Yeah, it's a, I was actually just talking about this on a previous interview I was doing, which I think is funny. I think there's a, there's a temptation to have like a really rigorous process for this. My personal approach is you have all sort of automations in two categories, and sometimes there's an overlap of the Venn diagram. Either number one, you are automating something that previously used to take you a lot of time to do. You are not changing the output quality of it whatsoever. You are just removing the manual steps in between. Outbound is actually a really good example of that in Clay today. Right? If I just take formulas enrichment and then I have my strategy of what I was Google searching, what information I was pulling, and then what text that relays into, I can just fully replace what was previously a manual system with an automated one. Now, I'm not changing the output quality of that whatsoever. It's gonna be the same exact thing versus there are also things that are purely just about adding new capabilities. So like a separate automation entirely is that something we don't do today because it is too manual, too tedious. Is everyone is sending really cool Clay tables in Slack from the customer team, from the sales team, from the marketing team. They're like, check out what this person did. And in the back of all of our brains, especially on marketing, we think we should feature these, we should talk to these people, we should build use cases around them, we should make templates. But that's nobody's core job. Everyone else has a different job and nobody actually wants to go copy paste those links with context into a database somewhere else to come back to 'em all at once at some point. I tried and, 'cause it only takes 3 to 4 seconds, right? It's a very quick thing to take someone's message and then send it to a database. But even that slight 3 to 4 second context switch is massively annoying. So we don't do it at all. Now with some of the AI agent technology that we have available, it's automatically running in the background. So I can just go in Monday morning and look at all the cool new Clay tables that got sent and read the updates, and now that is creating new possibility for marketing material that we didn't have previously. It's creating new possibilities for us to think about what new skills we want to teach, what people we wanna feature in our content. And then of course there are some things that naturally fall at the intersection of the two. The weekly updates thing that I just talked about is a great example of that. It saves time from people manually inputting that information, but it's creating new things that we were previously doing or not able to do because of that better information. So I think about first, which bucket the automation falls in. That's one part of the framework. The second part of the framework is time to value. How quickly can I stand something up? And that's a really easy one because either you're right or you're wrong about it if you're really honest, right? Most of the time I'm like, can I set up something that gets me 80% of the way there in 5 to 10 minutes? And if the answer is, I think yes, then I will take 5 to 10 minutes because that is a very low cost for something that could have a massive impact on us as a team. And if I don't get it done in 5 to 10 minutes, we benefit from Clay of the structure of we do have a whole team of GTM ops people now who are full-time jobs is figuring this stuff out. My background is in automation, so even if I don't get it perfectly there, I know what it would take to get it there and I can have someone else help me work on that. Or some of the technology is so good that it really does work in 5 to 10 minutes. The custom agents I just described for the table building automation and the weekly meeting notes all got set up in like 15 minutes with the Notion custom agent software. I'm pitching other companies, not even Clay now.

Curious, what's the rough estimate for how many companies you personally have gone in and helped over the last few years set up these processes, think through where to, um, where to apply automation? It's gotta be massive.

Yeah, I mean, if you're, if you're defining it just as like companies I've had these types of conversations with, whether or not I've worked with them formally, it was hundreds. Hundreds at this point. Yeah.

Hundred, right? Looking at what kind of patterns have you seen in terms of, you know, which companies are really gonna be able to take advantage of this and with a little bit of help be off to the races and which companies struggle, whether it's industry or organizational design or internal talents.

That is a really interesting question. I think there's a couple of different, once again, like elements to this answer, right? So element number 1 is size and scale of company. I think we all know the age-old adage of larger the company, harder it is to turn the ship around. Uh, that has been historically relatively true, right? Like the larger the company, the more established the division is when I've worked with them. Even if you have all the ideas, the irony is the problems are much clearer and easier to automate in those companies because they've been doing it the same way for so long. They have documentation and background. And so the A fixed world problem makes it a much easier thing to say, here's what you need to do to fix it. But the resistance internally and culturally is far higher, typically, especially because the number one thing I see with those types of companies, people are afraid that if you automate what they spend 30 to 40% of their day doing, that that means they get fired. Uh, and that's generally speaking the whole automation AI trend. Like my quick side note on this element of the answer is that if you look at history, and I still think this is true today. You will only see job capability continue to increase for the people who are willing to upskill, reskill, and get after it because the possibilities are always expanding. There are new careers. You could be a professional vibe coder today, and actually you don't even have to be employed by an organization today. Another side note that we can come back to is I think we will probably enter the most entrepreneurial generation of the past couple of decades because you don't need all the things or the obstacles that that were in your way of starting a business before. I am so excited about the like multiple 2 to 4 person shops, course creators, brands like you guys that are coming out of this era of AI. So one is size. The other two are going to be how automation literate or technical the organization is. And I'm gonna caveat that by saying, when I say automation literate, everyone assumes I mean technical skills. So when the people are hiring for automators, they think, do you know Python? Do you know Java? Do you know Zapier? Do you know Airtable? That's the same approach as having a checklist for, uh, if you're like going out and dating and be like, I want someone who's 6'1", who works in finance, who has blue eyes. You're like missing the forest for the trees. Maybe you have a type, but also the more deep thing that matters is do you have a connection? Do you understand the problem at play? Are you both on the same journey? What is actually really important to these companies early on is someone who knows how to think about automation. And that's actually the same thing with Clay and all these other AI tools. If you meet someone who is really good at just using Claude, but for whatever reason you put them in ChatGPT and they can't figure out their head from their feet, for lack of a better word, that's a red flag, right? Like that person does not have transferable skills. And in the world that we live in today, transferable skills are like the biggest way that I identify what I consider to be top talent or organizational change, because the skills you build today are not going to be the skills you need one month from now at the pace of AI evolution and change. And so that kind of bleeds into category number 3, which is sort of a continuous change culture and a willingness to do things different. That is most often correlated with category 1, which is size of company. The earlier you are, the more typically willing people are to do things differently. Depends on who the people are. But a good example of what a company that is like a medium size that does this well, I think is actually even Clay. Right? Like I was just talking to Varun, one of our co-founders, about I don't spend as much time as I used to playing with all the latest technology because I have so many other things that are going on in the team. But I still am spending weekends and nights 'cause I'm just too curious. Like I got into ClaudeBot, or it's Claw Open, Open Claw now, right? It keeps changing names. Multibot last week, and I didn't get to go fully deep with it because I didn't get it fully set up, but even from playing with it, without having to learn all of it, which comes back to category 2. I understand exactly what the potential is. I don't even have to have finished setting it up to know where it fits in my mental diagram of what is possible. And actually that's a really big leverage effect. Now I know what I think is possible. I can actually go find people who are experts at doing that thing and ask them to help me set up a small POC. So now we have someone that I know really well who's who's been doing this for a bit, uh, helping Varun's EA set up a Claude bot that will automate all of her, like, logistical EA things. And I was just actually, before jumping in this call, looking at some of the early setup. They took one day to do it. It's insane. It's automating so much stuff for her.

That's actually a perfect segue because we, we're trying to separate hype from reality here in, in this podcast. And Claude code is peak hype right now. What kind of work is moving over to Claude Code? Is it net new? Is it work that other tools were doing? For the uninitiated here, what's the deal with Claude Code?

Yeah, so if for the uninitiated, it's also probably worth noting that there's a lot of different things that all sound the same, right? There is Claude Code, C-L-A-U-D-E Code. Then there is also Claude Cowork, which is basically a more user-friendly version of Claude Code for your desktop. Claude Code, because it's in your terminal, feels more intimidating to people, but it is also technically more powerful than Claude Cowork because of what it has access to. Therefore, it's also a little bit more dangerous. You might accidentally delete very important files on your system. So proceed with caution. And then there's actually what is really, really hype right now, which is ClaudeBot with a W, C-L-A-W-D. I think they changed their name because they were probably getting infringement pressure from Anthropic about Claude. So now it's called OpenClaw. Which I wonder if they'll get OpenAI infringement on that. But, uh, it's funny because all that is is a local—

all—

it's a local agent you can install to your computer. And then it actually pops up a GUI, which is a graphical user interface that you can interact and chat with it through. And you can give it a bunch of skills, sort of like the MCP connectors that you may have tried in OpenAI or Anthropic, but it is running 24/7 constantly on your computer so long as it is open and it is figuring everything out. And you can interact with it via like Telegram or WhatsApp. So as an example, and I haven't been able to do this with MCP connectors for whatever reasons, actually I have some reasons. We can get into the technical details if we want to, but, uh, I'll, I'll give an example of what Tess was just showing me, right? Tess is Varun's EA. And so Varun has dozens of meetings every week and needs like a meeting brief for each of those people so he understands who he is talking to and why and what's going on. In a pre-AI world, Tess would just be manually collating all those things. God save her soul. In a post-AI world, we do have like an AI meeting briefer doc that we've created in Slack. So you can just go in and you can ask the bot, hey Dust, can you find me all the relevant customer information about John Adams and figure out what our past relationships are, who's been talking to him, what the current priority is. And because that information is stored across Slack, Notion and Salesforce, we can aggregate and pull it all into a really clean doc that then gets put manually from Tess into Varun's calendar. And now with this local agent that's powered by, you can use any of these AI models, you can use Grok, OpenAI, Anthropic, Gemini, whatever you want, uh, and give it access to these tools. It can go into Varun's calendar, find every single meeting he has, return to her a list of all the meetings that are on there. Then go into Slack for her, create the meeting summary docs for each of those people by asking the chatbot, take the links for each of those docs and put them in the calendar invite on Varun's calendar for each of those meetings. So all Varun has to do now is just go click on his meeting cal invite, look at the meeting brief doc, and all Tess has to do is sit back and look pretty while she focuses on other creative EA work, right? So like that is possible, like that is already done. The tough part, given the stage of where we're at, is that that is not a 5 to 10 minute setup. Like I, that one takes like 30 minutes to an hour if you're already deeply experienced with it. Uh, and it is a little bit more technical at the moment. So accessibility is lower, but this comes back to my point about as things get more and more accessible, more and more people will do it.

Can I chip in like real quick? And because I've been following, uh, like the evolution of the various names of, uh, OpenClaw and, uh, you know, I've It's on my to-do list to actually like set it up for myself as a one-person business. It's probably like something that you cannot pass on to. I want to also like play devil's advocate and get your thoughts because from what I see, obviously, like the real discourse on these type of things is on X, is not on LinkedIn. And that's where, you know, like the alpha is and where like all the, yeah, the early adopters and like I've read a lot about the risks of lack of control or potentially like this, like you effectively need to give it access to everything or nothing. And so how are you thinking about the drawbacks or the risks of giving too much control and autonomy to agents, especially when it comes to OpenClaw and What had, what are the steps that you're taking, you know, like, especially when it comes to, you know, like things like managing Varun's calendar and sensitive data that are in there?

Yeah. So I think there's, there's two elements of security risk that come to these types of tools, right? Like number one is just like your data and how is that data being used? So I think the APIs that you use, I'm blanking on the name that we're currently using to power the router calls for CloudBot internally at Clay, but it's very safe. It meets all the like ISO. GDPR, I'm not a security guy, right? So it meets all the standards. And so supposedly from a data perspective, meets our security standards that are pretty rigorous. So that part I feel pretty good about. If you're using the right tools, I'm happy to figure out what the tool is actually called in the follow-up. My friend who is the pro at it is actually the one doing it. And then number 2 is, I think, more important because this is what impacts more people, to your point, is if you give it too much access to tools, even if your data is secure, all of a sudden I did end up doing this for a little bit, right? Like this is part of the trial and error of setting it up. You have an AI bot responding for you and people are like, what? What did you just say to me? Uh, for example, I actually saw someone else do this to me as well. Like I got 3 or 4 emails to reschedule a meeting from someone that I had an interview with next week. And then they sent me a follow-up like 2 days later being like, I'm so sorry, that was my Claude bot just going in and like getting a little bit overeager. And so yes, that is a risk. I think some of that is going to be trial and error naturally. Like that is the cost of evolution in technology. And also the devil's advocate to your devil's advocate is that you can control scopes and permissions for each of the tools that it has access to. So for example, if I'm giving my Claude, let's call it OpenClaw, sorry, I'm just so used to seeing it on X to your point as ClaudeBot. Uh, let's say I give OpenClaw access to Slack. I can give it different scopes permissions. I can give it read permissions only. I can give it process permissions only. I can even control if it can only read messages in channels or in privates or in DMs and take away write permissions. And then the way that I'm prompting it right now, like as another example, is I had it go through all my Slack messages, summarize the most important themes for me, and then link back to the channels where those are if I wanted to go back for it. And then take all my DMs, keep them unread, but then draft responses for me so that now when I go through Slack, I just check all the DMs and then I just hit enter if I like it, or I just change it if I don't like it.

And you can manage your gas. So you can also manage conflicting objectives or priorities and stuff like that.

Yeah, it's all configurable, right? Like that's the, that's what makes it, this is like the trend in general, right? Like as everything moves more and more towards everyone has the ability to code, what we are losing, which was nice to lose in the intermediary, is all of the like really strict guardrails and like tightly defined ways we had built software to be used so that it was essentially quote unquote idiot-proof, right? But now, because the ability to code is more and more idiot-proof with all these different, and like, by the way, I'm not calling anyone an idiot. I consider myself the biggest idiot of all. Like now I am able to do all these things, right? Like I am finally able to go in and like prototype and stand up these things because before idiot Josh was unable to do it. And so. That comes with your greatest strength is also your greatest weakness, right? There's too many configurable options. You could get lost down the rabbit hole of trying to automate every single last step of what you're trying to do. And the biggest recommendation I would have there is Pareto principle is always in effect. Take the 20% that you can automate to give you 80% of your returns on the things that you need. Do that for as many things as you can before you start to try and optimize the last mile of your automations.

Right. We went pretty deep there on Claude and OpenClaw. That's amazing. I've got a lot to play with as well on that. Why don't I just switch gears for a minute? I mean, I've been seeing just this continual onslaught of innovation from Clay. What is the latest, Yosh?

Yeah, so there's so many things that are the latest, but probably the two that I am most excited about are You know, Clay's goal is to help you grow your business. Incredibly vague because there are so many ways that you can grow your business, but also really powerful because we are going to help you do it in every single way you possibly can. We started with outbound, we've done data enrichment, you've hopefully seen account-based marketing. The biggest addition to that onslaught of features has been ads. So historically, people were sort of able to duct tape together data enrichment, match rates, and then ad targeting mechanisms with HTTP API calls in Clay. We have now launched for enterprise-only customers, uh, ads across LinkedIn and ads across Meta that we can enrich data. And I think like triple to quadruple people's match rates on the targeting of who they're looking to get with these ads, because we're able to find data about the audiences that you're going after. So that one is really exciting because it is just such a black hole and black box for most people. That us being able to make an order of magnitude more impact for people's ads because we collect all the best data is a huge win for people running ads everywhere. We don't yet have Google, but it's coming soon. Um, and then the other one that is still in beta, so it's a little bit private, but I'm gonna keep teasing it cause I think it's now time to at least talk about teasing it, is what we are calling audiences. And so audiences, for anyone that has ever worked with CRM data, you know how painful it is to work in a CRM without Clay. And you probably also know if you've gotten this experience, how painful it is to work with a CRM. In Clay because it is still basically the closest you can get to programming in Clay without having to program. You have to pull in the original column, you have to add the conditional runs, you have to update different fields, you have to make sure you don't overwrite an existing field with a blank field by accident. You have to have a null handle exception for it, and then you have to update only the rows that actually have new data if you don't want to overload your API calls. At least you're not doing all that in code. But doing all that in clay is still relatively annoying. And so we have a much, much better interface for thinking about how to bulk handle data across your entire CRM that we have been testing with a couple of our early enterprise customers that has been going very well and that we are confident enough about that I can tell you it is coming to everyone else soon in the months to come.

I really can't wait for that. That's going to be awesome.

For sure.

Let's take a quick step back for the folks that are really at the beginning of their journey with clay. Uh, what are some of the biggest obstacles to getting going with it? And what are some tips to help somebody through it?

What are the biggest obstacles for someone who is new to the journey and learning clay? The answer to that question, I think, depends on who you are. I think broadly speaking, I have seen 3 big personas, and there's infinitely more, but these are the 3 most common ones that I see. Number 1, you are someone with a GTM background, and by GTM, I mean specifically GTM, not ops. So I'm taking out RevOps, sales ops, marketing ops from this. You're like a salesperson, a marketer, a field marketer, and you have heard about this Clay thing. You wanna automate more of your job. Let's go. Let's get into it. That person jumps into Clay. Is immediately lost, tries to do a couple of things, burns all their credits, gets frustrated, and leaves. Because 99% of the time, what that person is missing that is actually the first step to learning Clay is just the ability to think about problems from a computational sense. If you're a voracious reader, my favorite book on this topic is Algorithms to Live By. It is basically accessible, like layperson's book that goes through all the most complex algorithms we've developed in computer science. And how algorithms are actually just informed by the lives that we live. Technically, everyone here is running algorithms in their brain every time that you do something. When you get up in the morning to brush your teeth, that's an algorithm that you've trained yourself to run and how you do it. Hopefully your algorithm is effective and you're not getting cavities, right? That's a separate thing. But we are just translating patterns we see in the everyday world into systems that we build. So anyone can learn how to think that way, but if you have not been trained to think that way before, it's a whole new world of learning how to think in systems. And if you couple that with also learning to speak a new language, with also learning to have new friends in a new community, you would of course get social overwhelm. So number one is to sequence the order of things in which you're learning. So computational thinking for that first persona is the most important thing to grasp. My biggest recommendation for most people is actually not to start in Clay if you're in that bucket. Go take Codecademy 101 in Python on everything you could do from variables to for loops to if statements. Shouldn't take you more than 2 or 3 days and you will feel far better equipped to at least speak the language of the world you are entering. And then we can take that to the other 2 personas, right? So persona number 2 is someone with an ops background that comes into a software tool like Clay. That persona branches typically either you're the type of ops person that has been doing it all sort of manually and effectively for a while, so you've never touched a Zapier, an Airtable, or N8N. And so you then still fall into the Persona 1 bucket, which is you need to think about computational thinking. But if you are an ops person who is well-versed in no-code and low-code automation tools, it is likely that you at least have some element of computational thinking. And then what you need to think about is workflow thinking. How do you then map what types of problems are really good for Clay and what types of problems are better solved elsewhere? and I think the FET framework really helps with that. It's like Clay is really built for these massive data problems. It is for when do I start with data that I need enriching that sort of adds a go-to-market element to it that then needs transformation with AI that then needs to be synced and actioned elsewhere for a team. That's different from I need to update my calendar, you know, every single time I have a call. You could do that in Clay, but it's probably not the best thing to do that in Clay. And then persona number 3 that are actually sometimes the hardest personas to work with Engineers who come in with a lot of confidence in their ability to learn something that is less technical than coding. And that also branches into two buckets. There are engineers who intuitively get it and they just kind of run with it. Great. I don't need to worry about you guys. You have all the resources you need to be successful. There are engineers who will then message me and say, I have a degree in computer science and this thing is so confusing. You have built the most garbage piece of software that has ever existed on the planet Earth. I can't believe you even are an existing company. I have gotten messages like that before, and that is a long journey of essentially educating people on the fact that just because you have a computer science degree does not mean that you know how to think computationally. A big way to test for this is what I was mentioning earlier. If you have someone that is really good at writing and building software in Python, and that person is unable to translate it into JavaScript, if you give them a couple of pointers and tips, it is likely that they have been given a lot of fish. They have never been taught how to fish. And so those people still need to learn how to fish if they want to be successful in the world of all these expansive tools that build on the same core underlying concepts. That was a longer winded answer than I expected, but yeah. Awesome.

No, awesome answer. And I have a new book to read. So we talked about Clay. Talked about Claude. I know, uh, you're a heavy user of Dust. Is there one other tool that you think people out there should be using, getting familiar with?

Clay, uh, OpenClaw, Claude, Skills, Dust, Notion custom agents. I really like Notion custom agents. The weekly standup automation I was describing and the Creative Clay table automation I was describing are both actually Notion custom agents. And actually the thing that I am Recently playing with, just because I am someone with a shiny object syndrome, is I have Notion meeting recorders now running on most of my meetings that then automatically sort into a database with my transcripts that then have a Notion custom agent that goes and organizes and processes and then decides if there are buckets of action items for me to follow up on. So sometimes it's content for a social media post, sometimes it is messages to another team member, sometimes it is insights for the product team. And so now I don't have to think about, all of those things quite as much. I think Notion is really good for most people to like do custom agents plus meeting recorders plus other workflows. My other really favorite call recording tool is Attention, uh, for like specifically I would say GTM teams. Like if you are a GTM team looking to utilize AI on your call recordings, Attention is a no-brainer.

Okay. And why, why those, why those two, uh, in terms of Notion custom agents and Attention stand out in, in a crowd?

Yeah. So the reason Notion custom agents really stands out is because of the ease of setup and quality of use out of the gate. So Notion Custom Agents are the thing that, you know, I have tool fatigue as well. I'm not impervious to it. I don't want to try a bunch of new tools every day when I still have 100 Slacks and emails to get back to. And so the candid story on Notion Custom Agents actually is that I wasn't using them. And then because we have a close relationship to the Notion team and to their enterprise team, they were like, can you please just try it and let us know what you think? And because I'm the Notion person, or one of the Notion people at the company that was relegated to me. So I was like, all right, I'll take 30 minutes and like, let's just see if it can do this. I was skeptical. I was like, I'm gonna give them a bunch of feedback on why it's not gonna work. 'Cause I've tried to build this myself 4 times in the past. And lo and behold, it worked in just like those 10 or 15 minutes, right? Like I connected Notion to Slack, I connected my Notion database, I gave it which channels I wanted it to go in and index, and then I had it run an example on the last week of data and it was actually 95% of the way there. I spent the last 5% fine-tuning the edges. And so because it's fast, easy to set up, and it's automatically indexed on Notion and Slack, which are probably your team's two largest sources of unstructured data that you are not structuring efficiently and you're just losing insights in, that's what makes Notion Agents so powerful. The Notion Agents for meeting recordings and slash or the attention for meeting recordings is just a branch on like where that unstructured data goes, right? So I think about the biggest sources of unstructured data that probably hold the most insights in our work lives. Are your knowledge base. So Google Drive, Notion, et cetera. Probably your meetings calendar, which is also part of Notion Customs Agent. Definitely Slack, all the like haphazard conversations you're having. And then of course every single live call that you take because there is just mountains of insight on every single conversation that you have that could be traced back to the rest of the team. And so if you're not necessarily looking for GTM-specific use cases and you have sort of adjacent random automations you wanna build, I think Notion and the Notion meeting records Meeting recorder is sufficient, but then in the GTM world, right, you can also do this with Clay and Gong if you wanted to as well. Right. But Attention, I think, has done the best job of— they have this like AI sales assistant slash bot that goes in and indexes your Slack, your call recordings and everything else. So you can get a master system of intelligence. They also have automated AI flows you can build that connect to each of your call follow-ups. And I don't know what they're doing, but for whatever reason, like call recording data seems to be harder to get really high quality. AI on. Maybe that will change as models get more intelligent, but they are still the best quality at indexing over-call data that I've found till date.

Awesome. Thanks for the deep dive there.

And it's about how you, I guess, also restructure, like, again, chat, calls, calendar, you know, like there's a bunch of unstructured, very fluid data, right? So I think like what attention did so well, it's about how you structured that and package it in a way that's very useful for sales and GTM use cases, right?

Yep, exactly.

Awesome. I mean, we talked about a little bit of skills, we touched on tools, patterns that you see in GTM engineering. Let's chat a bit about talent. Everyone is doing everything. Everyone's expected to do everything, you know, like things are moving at more breakneck speed than ever, and the market doesn't seem to stop accelerating. I guess it's becoming harder than ever to stay relevant and to actually know what to focus on, right? So again, from your point of view, again, being at the forefront of this and this evolution of GTM engineering, you know, like, what do you think are the actual highest leverage skills that operators should apply in practice this year? I think we talked about, like you talked a lot about agency, but is there anything else or what would you say, what would you advise our listeners?

Yeah, that's a good question. And maybe my answer will be surprising. I actually don't think like technical skills or technical there's a level at which you need to get your technical skills and capability to, to stand out as a blank insert engineer of any sort. But that is really related again, more to the computational thinking, the workflow thinking. I actually think the further you want to get in that industry and career, the better it is for you to just be able to think about what is possible with relative accuracy and then be able to enable other people to build. Like my, and maybe I'm biased by my own example there. Because I don't spend as much time in these tools today anymore, right? Like I have a team and like now my priority is really like helping the team get better at doing all these automations. But in many ways, what has surprised me is just how much I am still GTM engineering, right? Like I am still sitting down with people on my team thinking about the problems and then they'll get stuck on something and I still understand, oh, here's how you would fix that. Here's the 3 things you should try. Here's who you need to go talk to and here's the tools I would try. That knowledge is what is actually the biggest unlock for getting people past stages where they get blocked. So I think just being able to sort of think in that way is still the most important thing. But once you hit that level of ability to kind of think and puzzle piece together problems, probably the next two important things that really feed into that ability are number one, drinking from whatever fire hose you choose, right? Like, yes, the information is getting faster. Yes, it's also constantly overwhelming. And also maybe the like solace I would offer to people is that the one thing that's not changing is the number of hours in a day and the things that you work on. And so a lot of the distillation of signal to noise is great. Shiny object syndrome. Amazing. Keep it shiny unless you have a real example of something you wanna do in your life. And if you're just being intentional about what you're doing in your day-to-day at work, you know what the things are that bother you and you know what the things are that inspire you. Focus only on those things. So if you're, for example, someone that doesn't have a lot of like administrative boring work in their day-to-day job or like the ability you think to automate anything that you do, which I would question, but like as an example, you don't care about Open Claw, right? Like Open Claw is not really going to do anything for you. It's like a nice to have interesting thing to learn about. Maybe if you make a career pivot, great. But like, listen to it, understand it, don't spend time on it. That's the signal to noise differentiation that I think is really important and an increasing velocity. Of information updating. And then the second thing, which is I think the real glue for all of it, is just being able to learn quickly, right? The transferable skills, sort of the adaptable skills is maybe even a better word for it. If you are really stuck in one way of thinking, viewing, or working in the world, you are probably most at risk for disruption because the world is changing so quickly. And so your ability to learn new things on the fly or to take complex new concepts and quickly distill them down into something actionable helps keep you ahead of the curve and helps also keep you at the forefront of what's possible.

It's super interesting because like you've moved from like being an IC, GTM engineer, automation type of role, you know, hardcore onto ClickTables. And now you have your task to keep the team together, you know, keep them productive, keep them moving forward. And how do you balance this? Like, this new set of skills, you know, like about more people management and keeping the team together, and at the same time, keeping your mind, as you said, upskilled, reskilled, fluent to new learnings and on what's next?

Yeah, I think it's a good question. And I think it sort of comes down to two core things. So thing number one is that I think this is true of like the Klei company culture in general. We don't really have a lot of like very pure managers. Like everyone who's in some sort of a management role is also a player coach in many ways. And so that means that you ideally are just as able to like get back into the weeds with the ICs that are on your team as you are able to have help strategically guide them. And so I think because of that, it relates to what I was saying earlier about in many ways, ironically, which I wouldn't have predicted, I feel like I am doing more GTM engineering thinking. Than I ever was as an IC because I'm thinking constantly about like, what are all the different problems and how would we solve them? And then I'm helping enable people to learn the skills that they need to go after it and do it. Right? So like, because I know what is possible and I'm very good at understanding like where the boundaries lie, every time someone is doing something, like the first person in the room to be like, oh, I know how we could automate that is still me. And then I know exactly what tools we would use. I know who to send them to. And then I say, go talk to these people. Go try to do it this way. If you hit an issue, come talk to me and I'll unblock you. And then I know what the issue is and I know how to fix it, right? But they've done a lot of the like setup and they're actually learning for themselves how to kind of piece together these different automations. So the player coach thing is one part of it. And then the other part of it is just, you gotta find time to like hack away on things that you're genuinely interested in. And I think I've just always been obsessed with productivity and automation. And so I felt behind the 8 ball. I think I am behind the 8 ball on OpenClaw relative to where I used to be, you know, a year or two ago. But that is maybe a feature and not a bug as well, because the big thing that has changed for me is my life, right? A year or two ago, I was just sitting on an island automating bananas falling and coconuts growing, you know? Now I'm not on an island. Now I have other people that I think we're also all kind of working together. And so I do need to pay attention to more of those people before I figure out what other things are fun to do on the island. And so I'm a little bit further along the curve, but I'm still at a place that works for me.

So out of these all meta skills that you mentioned, I can, I'm taking away agency, workflow design or systems design and thinking, curiosity, learning in and on itself. So like hands-on. I've been reading and hearing a lot more about taste as well, which I, in my own, you know, like dictionary, I define it as the ability to discern what's good, what's bad within a specific context. Right. What's your take on this? How do you define it and how do you train it?

Yeah. So one, I agree. I think taste is becoming ever more important of a skill. I don't know if taste's important in our culture and society has changed necessarily or importance. I think it's just that because The amount of average to slightly above average stuff is increasing. Taste is becoming more evident, right? I think it's always been the case that really, really high quality things stand out from the crowd. And high quality, I think it's important to mark, is not necessarily, or high status or high prestige, is not the same as having taste, right? Taste is really just about the pureness of the craft itself. And so you can apply that very general framework to any domain that you want. So if we apply it to software and GTM, it is— taste is purely about in a software that's built to make you more efficient, how good is it at making you more efficient? How could it be better at making you more efficient? And how does it compare to everything else that's available to you out there? If it compares favorably on all those dimensions, then that's something that is built with very high taste. Now, if you compare taste to like this other category, I would say of everyone's afraid that vibe coding software will eat up all these traditional industries. Like the famous one that I saw on X a few months ago was Dropbox, right? Like Dropbox is— I, or sorry, not Dropbox, DocuSign, legal signatures, right? Electronic signatures. I can vibe code a software that then I think DocuSign sued this person for. And taste is a little bit different there, right? Because you're fulfilling the form or objective as easily as you want. And so taste maybe doesn't matter as much, but also the reason DocuSign— again, taste is dependent on the person. The reason DocuSign is not disappearing as a company overnight is because there is a lot of legal minutia, security, uh, risk, and company trust that gets built around very important legal documents. And so if you are someone that is really deeply entrenched in that field and you have high taste, you're still sticking with a company that provides all of that for you. I think it's just that a lot of people who don't have taste, or taste is maybe not important to them in certain dimensions. Like if you are a small business owner, legal taste probably does not matter to you. Just get the deal signed and move on. So what matters to who will change. And then I think to answer your last question, the way you develop taste is just by spending a lot of time repeat exposing yourself to things in that industry. So for example, I don't really have a lot of legal taste, nor do I have a desire to have a lot of legal taste because taste comes at an opportunity cost. You can only develop taste for so many things. I have a lot more taste for really high quality books, for automation software, which is a weird thing to have taste for, but I guess is a thing. For, I play a lot of racket sports. I have very strong opinions on like what makes a good tennis player versus a bad tennis player, even if someone wins versus doesn't win, right? And like that ability to nuancedly distinguish between outcomes, quality, fit, and relevancy is, I think, the layer of taste that you need to get to. And you can only get to it through experience, exposure, and repetition.

I love the opportunity cost element to it. I never thought of it. So thanks so much for sharing.

Sure. Thanks for asking it.

All right. I think we're probably coming close to the end of this episode. I think we've had so many great conversations on this podcast, and I think this was absolutely one of my favorite. Uh, Yasha, really thrilled that you're able to chat with us today. We're thrilled that you're going to be continuing with Go-To-Market Engineer School as one of our lead instructors. Um, just to wrap up today's conversation, we, we really unpack what go-to-market engineering actually looks like in practice. From— you talked about moving grunt work up a layer of abstraction with AI, uh, and, and building systems that, uh, let teams think more like engineers but without ever having to touch Python. And we dug into how to prioritize what to automate, uh, what tools like Clay, Claude, Open Claude, Notion custom agents, and Attention are what they're actually good for. And we talked about the mindset shift that operators need to stay relevant as, as this stack continues to evolve. I think finally, if there's one takeaway from this conversation with you, Yash, it's that the people who learn to think computationally about their workflows and then keep iterating, these are the ones that are, are gonna compound the most value over the next few years. Agreed.

Great result.

Yeah. So thank you, Yash. Um, uh, this was fantastic. Uh, any, any place, uh, besides LinkedIn where you, where, where our audience can, can catch you?

You can catch me on LinkedIn or you can catch me on X. Uh, also love jamming with people on Threads there. Nothing else immediately comes to mind, but Appreciate you guys having me as always. Always love jamming and chatting and I'll see you guys in the next one.

Thanks.

Have a great night.

See you guys.